OpenLabeling: open-source image and video labeler

Image labeling in multiple annotation formats:

• PASCAL VOC (= darkflow)
• YOLO darknet
• ask for more (create a new issue)...

Citation

This project was developed for the following paper, please consider citing it:

@INPROCEEDINGS{8594067,
  author={J. {Cartucho} and R. {Ventura} and M. {Veloso}},
  booktitle={2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  title={Robust Object Recognition Through Symbiotic Deep Learning In Mobile Robots}, 
  year={2018},
  pages={2336-2341},
}

Latest Features

• Jun 2019: Deep Learning Object Detection Model
• May 2019: [ECCV2018] Distractor-aware Siamese Networks for Visual Object Tracking
• Jan 2019: easy and quick bounding-boxe's resizing!
• Jan 2019: video object tracking with OpenCV trackers!
• TODO: Label photos via Google drive to allow "team online labeling". New Features Discussion

Table of contents

• Quick start
• Prerequisites
• Run project
• GUI usage
• Authors

Quick start

To start using the YOLO Bounding Box Tool you need to download the latest release or clone the repo:

git clone --recurse-submodules git@github.com:Cartucho/OpenLabeling.git

Prerequisites

You need to install:

• Python
• OpenCV version >= 3.0
  1. python -mpip install -U pip
  2. python -mpip install -U opencv-python
  3. python -mpip install -U opencv-contrib-python
• numpy, tqdm and lxml:
  1. python -mpip install -U numpy
  2. python -mpip install -U tqdm
  3. python -mpip install -U lxml

Alternatively, you can install everything at once by simply running:

python -mpip install -U pip
python -mpip install -U -r requirements.txt

• PyTorch Visit the link for a configurator for your setup.

Run project

Step by step:

1. Open the main/ directory

2. Insert the input images and videos in the folder input/

3. Insert the classes in the file class_list.txt (one class name per line)

4. Run the code:

5. You can find the annotations in the folder output/

   python main.py [-h] [-i] [-o] [-t] [--tracker TRACKER_TYPE] [-n N_FRAMES]
   
   optional arguments:
    -h, --help                Show this help message and exit
    -i, --input               Path to images and videos input folder | Default: input/
    -o, --output              Path to output folder (if using the PASCAL VOC format it's important to set this path correctly) | Default: output/
    -t, --thickness           Bounding box and cross line thickness (int) | Default: -t 1
    --tracker tracker_type    tracker_type being used: ['CSRT', 'KCF','MOSSE', 'MIL', 'BOOSTING', 'MEDIANFLOW', 'TLD', 'GOTURN', 'DASIAMRPN']
    -n N_FRAMES               number of frames to track object for
   

To use DASIAMRPN Tracker:

1. Install the DaSiamRPN submodule and download the model (VOT) from google drive
2. copy it into 'DaSiamRPN/code/'
3. set default tracker in main.py or run it with --tracker DASIAMRPN

How to use the deep learning feature

• Download one or some deep learning models from https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md and put it into object_detection/models directory (you need to create the models folder by yourself). The outline of object_detection looks like that:
  • tf_object_detection.py
  • utils.py
  • models/ssdlite_mobilenet_v2_coco_2018_05_09

Download the pre-trained model by clicking this link http://download.tensorflow.org/models/object_detection/ssdlite_mobilenet_v2_coco_2018_05_09.tar.gz and put it into object_detection/models. Create the models folder if necessary. Make sure to extract the model.

Note: Default model used in main_auto.py is ssdlite_mobilenet_v2_coco_2018_05_09. We can set graph_model_path in file main_auto.py to change the pretrain model

• Using main_auto.py to automatically label data first

  TODO: explain how the user can

GUI usage

Keyboard, press:

Key	Description
a/d	previous/next image
s/w	previous/next class
e	edges
h	help
q	quit

Video:

Key	Description
p	predict the next frames' labels

Mouse:

• Use two separate left clicks to do each bounding box
• Right-click -> quick delete!
• Use the middle mouse to zoom in and out
• Use double click to select a bounding box

Authors

• João Cartucho

  Feel free to contribute